and required user or PAM permission. The handover as well as the corresponding pickup process may last additional time. In comparison to the barrier system’s check-in times of around 7 to 10 seconds, this may mean that several handover and pickup zones will be needed to accomplish appropriate handover and pickup times especially at rush hour. Timp-ner et al.109 shows that pick-up times of about 1 min are possible. Thereby, the availability of the AVP service has to be given. The separation of entrances for manually driven and automated vehicles could be considered to avoid annoyed customers. Even in case of sepa-ration, AVP systems have to provide value in terms of comfort and time savings to receive the acceptance of customers. This also may result in appropriate ways for service payment which avoid waiting queues, e.g. by paying via app rather than using a conventional ticket machine. The functionalities of a typical terminal found in today’s parking garages need to be extended. An intuitive HMI concept is required for the communication between users and AVP systems to instruct handover/ handback requests, to retrieve the current vehicle status or to pay service.
Automated driving to a point of interest: When driving to a point of interest, the infrastruc-ture may support optionally. A static map of the corresponding parking facility can be stored and transferred optionally at the entrance. An integration of device storage and C2I modules into the infrastructure will be necessary for the transmission of the digital map. Additional infrastructure sensors are required for the parking space occupancy status if these are not already integrated. The occupancy status in today’s parking garages is visualized with optical signals for drivers, but needs to be transferred to automated vehicles. If perception modules are outsourced into the infrastructure, the parking garage has to be equipped with environ-ment sensors for object state estimation, especially at locations in which occlusion is una-voidable such as ramps. As a result, planning and perception modules will require additional processing units for signal processing and C2I modules for data transmission. Occurred blocking which may be solved easily by human drivers e.g. by performing evasive maneu-vers or tight maneuvering, become more challenging for automated vehicles e.g. if evasion is prohibited by safety design. Deadlocks formed by automated behavior have to be pre-vented to ensure the availability of the overall parking service. A human supervisor may support in terms of troubleshooting such as blocked automated vehicles, collisions, break-down of function modules or activation issues. Limitations in indoor positioning110 or in measurement accuracy may restrict vehicle types simply due to narrow parking construction.
The results of the MRP and MRS zone show that restrictions of executable maneuvers or restrictions in velocity may have to be introduced for cost-efficient AVP systems.
Automated maneuvering into the parking space/ Automated leaving of the parking space:
When parking the automated vehicle, an optimal utilization of the parking area is desired for
109 Timpner, J. et al.: k-Stacks: High-density valet parking for automated vehicles (2015).
110 Einsiedler, J. et al.: Vehicle indoor positioning: A survey (2017).
capacity and throughput purposes, especially for peak demands. High density parking111,112 is a promising solution for the increasing demand of parking spaces in urban areas by pack-ing vehicles denser since humans do not need to access automated vehicles in the parkpack-ing area. Ferreira et al.113 presents parking layouts which reduces 50% of the necessary parking space compared to conventional parking. However, high density parking becomes challeng-ing in a mixed traffic since vehicle doors should be openable. This may result in a separation of existing parking garages in two parking space blocks to enable high density parking in a mixed traffic and to ensure optimal utilization of parking area.
Further issues:
The advances towards electromobility114,115 still have two major disadvantages: reduced driving ranges and increased charging duration. AVP may ease the traveler’s transfer by charging electric vehicles while these are parking. Consequently, charging stations to pro-vide additional service have to be integrated in today’s parking facilities either for each park-ing space or by sharpark-ing chargpark-ing stations. Electric vehicles are switched once the chargpark-ing process has finished. The switching process has the advantage of less required charging sta-tions. A docking process of automated charging to serve different vehicle types has to be established. Finally, a legal basis is required in case of caused harm to participants in the parking garage. A legal basis is required to clarify the responsibilities between manufacturers which design AVP systems, parking garage operators which provide AVP systems and cus-tomers which use the systems. The issue becomes especially crucial if a cooperation between infrastructure providers and manufacturers of automated vehicles takes place.
The discussion shows that a major issue for today’s parking garages will be a quick handover and pick-up process at rush hour to ensure time savings. Additional efforts lie in the preser-vation of service availability, the infrastructure support of automated vehicles, the imple-mentation of high density parking, the integration of e-mobility for AVP systems as well as the provision of a legal basis. The derived AVP configurations and minimum criteria ease the migration of AVP systems in today’s existing and newly constructed parking garages by integrating the necessary safety design for a preferred AVP configuration. Parking garage operators and manufacturers can influence their required degree of infrastructure support by maneuver or vehicle type restriction. Both can choose between personally preferred version of AVP configurations based on the relevance of costs, time-efficiency, safety, availability and accessibility.
111 Timpner, J. et al.: k-Stacks: High-density valet parking for automated vehicles (2015).
112 Banzhaf, H. et al.: High density valet parking using k-deques in driveways (2017).
113 Ferreira, M. et al.: Self-automated parking lots for autonomous vehicles based on vehicular ad hoc network-ing (2014).
114 Schwesinger, U. et al.: Automated valet parking and charging for e-mobility (2016).
115 Klemm, S. et al.: Autonomous multi-story navigation for valet parking (2016).
9 Conclusion and Outlook
Automated Valet Parking (AVP) systems are one of the first systems in automated driving that may be introduced soon by manufacturers116. The race of the development has already started.117 However, a clear definition of minimum criteria for AVP systems is crucial to ensure safety by design in the early development process. The minimum criteria shall hereby consider diverse parking garage topologies and the required allocation of responsibilities between the infrastructure and the automated vehicle for a safe AVP service. The integration of minimum criteria in the early design process shall hereby minimize the risks of harm.
In particular, the state-of-the-art reveals a lack of minimum criteria for the use case AVP.
Today’s standards assume for AVP the minimum capability of longitudinal and the lateral control performance, the monitoring of the driving environment and a minimal risk state in case of a fallback situation. None of the existing literature describes how to integrate AVP into existing diverse parking structures and which necessary conditions are expected from cooperative AVP systems to prevent risks of harm. Today’s technical realizations focus on the provision of a digital map at the entrance in combination with the detection of free park-ing spaces to assign a free spot. However, additional beneficial AVP configurations are pos-sible which were not yet addressed in the state-of-the-art. Two major research questions (RQ) were derived in this thesis:
RQ1: What is the essential subset of minimum criteria AVP configurations require to fulfill for safe operation?
RQ2: Which degrees of infrastructure support are needed and what are their benefits?
The contributions of this thesis regarding the identified research questions can be mainly summarized as follows:
First, this thesis decomposes the AVP service into functional scenarios for system abstrac-tion.118 Functional scenarios are used to give a functional description of the system. Major scenarios are the vehicle handover to the parking area management system, automated driv-ing to a point of interest, automated maneuverdriv-ing into the parkdriv-ing space, automated leavdriv-ing of the parking space, vehicle handover to driver and aborting the valet parking procedure.
Identified functional scenarios serve as an input for the followed derivation of safety require-ments from safety goals, the specification of a minimum required perception zone and the identification of function modules.
116 Gasser, T. M. et al.: Rechtsfolgen zunehmender Fahrzeugautomatisierung (2012). pp. 7-8
117 Banzhaf, H. et al.: The future of parking: A survey on AVP with an outlook on high density parking (2017).
118 Schönemann, V. et al.: Scenario-based functional Safety for AD on the Example of AVP (2018).
Second, this thesis introduces the derivation of low level safety requirements from safety goals in compliance with the international standard ISO 26262 and SOTIF.119 The hazard analysis and risk assessment has revealed that an unintended activation of the valet parking function outside of the infrastructure-controlled parking area and incorrect data transmission between parking area management are categorized as most dangerous. Moreover, a key fac-tor is to avoid collisions between the ego-vehicle and other objects such as pedestrians or manually driven vehicles. Thereby, the object’s state variables, its existence and class have to be determined to successfully avoid a collision. The state variables include the object’s dimensions, the object’s pose and the object’s velocity. A threshold is defined for the maxi-mum allowed longitudinal and lateral measurement error of automated vehicles.
Third, this thesis presents the mathematical and geometrical formulation of a minimum re-quired perception (MRP) and safety (MRS) zone.120 In particular, a definition of an area, in which the determination of the object’s state variables, its existence and its class is manda-tory for collision avoidance, is given. The magnitude of this area is maneuver-specific and therefore an investigation of occurring maneuvers for each individual parking garage is re-quired. The maneuvers following a straight lane, driving backwards, crossing an intersec-tion, turning left/right have been investigated. The worst case for the resulting stopping dis-tance is formed when a manually driven vehicle moves with the maximum allowed velocity towards the ego-vehicle and both vehicles are breaking. In case of turning left/ right the vehicle covers a tractrix curve. The superposition of the ego- and object’s stopping envelopes at maximum allowed velocities for the executable maneuvers in the operational design do-main forms the MRP zone. A reduction of the MRP zone is possible by introducing auto-mated driving only or by restricting executable maneuvers in the parking garage. The MRS zone, a subset of the MRP zone, determines the last possible border in which a deceleration requires to be triggered. However, the MRS zone has its limitations. The MRS zone assumes a linear moving behavior of the object due to the knowledge lack of its intention. A prototype of the MRS zone shall present the feasibility and applicability for collision avoidance (proof of concept) in the context of automated valet parking, but requires additional extensive test-ing.
Fourth, this thesis illustrates the functional modules to derive needed AVP configurations.
Hereby, the derivation of needed AVP configurations required the identification of function modules of an AVP architecture. In the first step, system requirements are assigned to func-tion modules. The funcfunc-tion modules form an AVP system architecture. The architecture is split into a three-level hierarchy of the driving task according to Donges121,122. At navigation level, a suitable route is chosen from an available road network by considering the road
119 Schönemann, V. et al.: Fault Tree-based Derivation of Safety Requirements for AVP (2019).
120 Schönemann, V. et al.: Maneuver-based adaptive Safety Zone for infrastructure-supported AVP (2019).
121 Donges, E.: Aspekte der aktiven Sicherheit bei der Führung von Personenkraftwagen (1982).
network. At guidance level, a behavior planner selects a sequence of behaviors that enables the vehicle to reach the assigned destination by taking into account other traffic participants or objects and the traffic regulations. A maneuver of the sequence is selected in the maneuver planner and is passed to the trajectory planner to determine a collision-free trajectory based on the maneuver specification. At stabilization level, the deviations between target and actual trajectory are evaluated and minimized via corrective control actions (closed-loop control).
Finally, this thesis demonstrates the distribution of function modules between the infrastruc-ture and the automated vehicle. The analysis shows that there is no optimal system architec-ture for the given impacts. A tradeoff exists between overall costs, time efficiency, safety, and availability of AVP systems with today’s vehicles. The analysis illustrates that the num-ber of derivable AVP configurations is too large for comparison on configuration level.
Therefore, this thesis presents the benefits of infrastructure support by gradually assigning function modules towards the PAM. A fully vehicle-based AVP has to perform the automated driving task without the help of an infrastructure (Configuration 1). Time-efficiency can be increased at minimum effort by transmitting a static map and the amount of free and occu-pied parking spaces (Configuration 2). Further efficiency can be established at the costs of environment sensors by detecting free and occupied parking spaces to assign free parking spaces directly at the entrance (Configuration 3). An increase in vehicle velocity triggers the need for an infrastructubased environment perception. Infrastructure sensors provide re-quired occluded areas (Configuration 4). If the PAM additionally takes over the route plan-ner, several vehicles can be coordinated for most time-efficient placement and congestion avoidance (Configuration 5). An infrastructure-based AVP takes over all perception and planning modules and sends the required control commands to the automated vehicle (Con-figuration 6).
The contributions of this thesis add value to the design of future AVP systems:
AVP configurations and minimum criteria ease the migration of AVP systems in today’s ex-isting and in newly constructed parking garages. Minimum criteria lay the foundation for the development of a necessary safety design. Manufacturers and suppliers benefit from the identification of minimum criteria by integrating the derived minimum criteria in their early system development process to ensure safety by design. Safety by design will minimize the risks of harm caused by developed AVP systems. Hereby, the safety by design affects the infrastructure and the automated vehicle. The fulfillment of minimum criteria can be achieved cooperatively between the infrastructure and the automated vehicle. This thesis illustrates the needed infrastructure-support for AVP systems. However, manufacturers have the option to distribute the functions between both entities according their preferences to accomplish the defined minimum criteria. Manufacturers and suppliers benefit from the specification of necessary conditions for AVP systems. The developers only require to con-sider the necessity for an individual parking garage and therefore save efforts by avoiding unnecessary above needed safety performance. Safety requirements are reduced by consid-ering the specific parking garage topology such as the executable maneuvers or reducing required stopping distances by parameter adjustment for the individual parking garage. For
example, the defined MRP zone can be used for the necessary sensor coverage for a maneu-ver-specific parking garage and the integration of required infrastructure support at specific locations such as ramps or occlusions. These safety critical spots can be equipped with in-frastructure sensors to increase safety and time-efficiency. AVP systems can be configured with the necessary sensor coverage by saving costs through necessity. AVP systems can be configured according the required safety performance for the individual parking garage. The MRP and MRS zone provide the option to control velocities dependent on the distance to occlusions and to traffic participants. The MRS zone can be integrated for collision avoid-ance. Parking garage operators are more willing to invest in AVP systems which provide less risks of harm to customers.
Parking garage operators benefit from AVP services with less risks of harm. As a result, less harm caused by AVP systems will ensure less collisions and therefore will positively influ-ence the availability and throughput of the parking garage. Customers will accept a safe AVP service more likely. Parking garage operators save costs by only investing in the required safety configuration for their specific parking garage. There is no need to pay for a more extensive AVP system. Furthermore, parking garage operators can restrict executable ma-neuvers, vehicle velocities or vehicle types to reduce safety requirements and corresponding costs for their infrastructure modifications. Layouts for newly constructed parking garages can be designed to ensure time-efficient and safe AVP systems. A safe AVP service provides the opportunity to integrate high density parking for the increase of parking capacity. The progression of e-mobility in combination with AVP systems allows the parking garage oper-ator to integrate charging stations in the parking garage and provide additional value for customers. Parking garage operators can select between AVP configurations according the constraints present in their parking facilities. They can prefer which distribution of functions is more beneficial for their individual circumstances. Some may favor a low cost application whereas others find time-efficient services more attractive.
Customers of the parking garage benefit from the reduced risk and a safer execution of the AVP service. They are more willing to use and pay for a safer AVP system which saves their valuable time and releases them from the burden of parking manually. A safer AVP system causes less harm to health and life of parking garage participants. Since the parking garage operator invested in a parking garage-specific safety configuration, lower costs of the oper-ator’s modifications are shifted towards customers. Safe AVP systems indirectly enable high density parking and charging stations. Customers may benefit from lower parking ticket prices due to high density parking. Parked electric vehicles of customers are charged during the parking process.
However, still many issues are unresolved in the AVP domain. Future work needs to be done to pave the way for automated parking pilots. This thesis proposes a three-fold strategy:
First, AVP systems have to provide value in terms of comfort and time savings to receive the acceptance of customers. This may mean to ensure a flowing traffic through the increase of handover/ pickup zones and provide solutions for a breakdown of the AVP system. Long
waiting times need to be avoided. The AVP system needs to be migrated in today’s existing parking garages.
Second, AVP systems need to be beneficial for parking garage operators and manufacturers.
AVP systems have to be economically. The commercialization of AVP systems requires cus-tomers willing to pay for AVP systems. This may result in high density parking which in-crease the demands on future AVP systems in terms of time efficiency, localization and ad-ditional constructional changes.
Third, legislative authorities need to approve AVP systems. This raises the question of lia-bility in case of accidents or fatalities. As a first step, the elaborated minimum criteria can be used as a checklist to ensure safety by design. Minimum criteria for AVP systems will also be required for liability reasons. A more complete set of minimum criteria is desired to additionally minimize risks of harm. The corporation of independent entities, operators and manufacturers is required to increase the set of minimum criteria and integrate them in the early system development phase. Minimum criteria need to be specified on low level instead of at overall system level. For example, maximum allowed lateral and longitudinal errors for state variables have to be broken down on parameter level to assign individual errors for the determination of position, orientation, dimension and velocity. Furthermore, a threshold for the existence probability to specify the accepted amount of false positive and false negative detections is desired. Further verification and validation of the minimum required safety zone and its applicability could be progressed. Hereby, the limitation in motion prediction for collision avoidance has to be addressed.
Figure 9–1: Decomposition of the automated driving system in functional scenarios which served as input for the specification of safety requirements from safety goals (left), a minimum required per-ception zone (middle), identification of required module functions (right) to derive minimum criteria for AVP and characterize possible degrees of infrastructure support
RQ2:
Which degrees of infrastructure support are needed and what are their benefits?
RQ1-2
Safety Requirements
Determination of Pose, Dimensions, Velocity, Existance, Class.
Maximum lateral and longitudinal Measurement Error.
Plan collision-free Trajectory.
Prevent unintended actuator commands.
Prevent unintended AVP-Activation.
...
Automated Driving System
RQ1-3 RQ1-1
Chapter 4/5 Chapter 6 Chapter 8
Chapter 8
Chapter 3 Item Definition: Decomposition in functional Scenarios
Pick-up Zone Parkin g Spot
Handover Zone
dreq,obj
dreq,egoF
dreq,obj
dreq,obj dreq,obj
dreq,egoR
Topological Map
World Model
Mission Planner
Route Planner
Behavior Planner
Maneuver Planner
Trajectory Planner
Mission
Target Behavior
Maneuver
Actuators
Self-Perception Environment
Perception
User
Handover/ Handback/ Abort
HMI Status
Trajectory Controller
+ Actual Control Value
Planning
Eigenstat e TrafficScenery Route Network
Route
Trajectory
Navigation LevelGuidance LevelStabilization Level Activation
Object State Estimation Modeling Local Scenery
AVP System
Sensors
Occupancy Status
Rest rict ion
Status Status Parking Space
Occcupancy Free/Occupied
Status Status
Control Loop
Target Control Value
Fu lly v eh ic le - b as ed A V P
Fu lly v eh ic le - b as ed A V P Fu lly v eh ic le - b as ed A V P Fu lly ve h ic le - b as ed A V P Fu lly ve h ic le - b as ed A V P
Fu lly v eh ic le - b as ed A V P
Fully vehicle-based AVP Fully infrastructure-based AVP
Config. 1
Config. 2 Config. 3
Config. 4
Config. 5
Parking Spot
Config. 6
RQ1:
What is the essential subset of minimum criteria AVP configurations require to fulfill for safe operation?